Are there long term architectural advantages that Titus will have over Kubernetes? Given K8s is multi cloud it seems like the investment there will continue to be very strong and will eclipse Titus, even on AWS.
Any high level reasons Titus will win on AWS long term?
As awesome as kubernetes is and is becoming, I've always been a fan of the Mesos architecture. For managing large, complex, diverse workloads, it's perfect.
Titus seems to be the cost of Netflix's brilliant engineering talent. They clearly had a need for container orchestration before Kubernetes was available and by this point they have sunk years of effort into their current investment in Titus.
At this point they have surely recognized that internal closed tools will never compete with the velocity allowed by a huge open source project like Kubernetes and they are left with two options:
* Recognize the sunk cost of Titus, avoid the relevant fallacy, and migrate to Kubernetes
* Make Titus open source and hope that enough community can be built around it to justify the long-term cost of continuing its use internally.
It certainly is possible Titus gains a lot of steam, particularly from orgs that are already on Mesos, and be a viable long-term option. However, if I was a betting man, I'd say that 10 years from now the world has moved on and even Netflix will be using Kubernetes, or the future iteration of container-orchestration orchestrator that will inevitably be built on top of it.
It also considers bandwidth requirements as I see, which Netflix surely needs and schedules containers accoridng to that, because of each network interface and instance has some limits.
Very excited to try this out finally. Integration with netflix OSS ecosystem seems like it will be a big win for moving those dependant services off ec2
* Overview: https://netflix.github.io/titus/overview/ * Prerequisites: https://netflix.github.io/titus/install/prereqs/ * Building Code: https://github.com/Netflix/titus-executor/blob/master/README... * ACM Queue Article: https://queue.acm.org/detail.cfm?id=3158370 * License: APL 2.0